Skip to main content

Microtechnology and integrated microsystems to investigate neuronal networks across scales

Periodic Reporting for period 3 - neuroXscales (Microtechnology and integrated microsystems to investigate neuronal networks across scales)

Reporting period: 2019-10-01 to 2021-03-31

SUMMARY
The proposed project constitutes an interdisciplinary engineering- and systems neuroscience-driven effort, aimed at a deeper understanding of the behavior of neurons or brain cells across scales. Across scales pertains to the spatial domain - from details of subcellular components through single neurons to entire networks - and the temporal domain - from single action potentials to long-term developmental processes. We use rodent cortical neuron cultures and brain slices, more recently also organoids and human induced pluripotent-stem-cell-derived neurons. The project methodology includes a combination of engineering, neuroscience and software/algorithm developments: microelectronics-based high-density microelectrode arrays for recording and stimulation, patch-clamping directly on the microelectrode chips, high-resolution upright confocal microscopy, genetic methods, large-scale data handling strategies, and the development of dedicated data analysis and modeling algorithms. We aim at studying - at the same time in the same preparation - details of specific neurons and subcellular components (somas, axons, synapses, dendrites) in their functional context and the characteristics of the corresponding networks (functional connectivity, emergent properties, plasticity). We will study alterations of components and networks over time and upon defined perturbations and mutual interdependence of network and component characteristics. Potential applications include research in neural diseases and pharmacology.

IMPACT AND IMPORTANCE
The project opens new pathways and possibilities to study components of neuronal networks in their functional context, and, at the same time, the corresponding networks, as well as mutual interactions and interdependence between components and networks. This project provides new insights into details and fundamentals of neuronal features and information processing. Examples include details of axonal signaling, or high-throughput monitoring of all action potentials of all neurons in networks over extended time to see developmental effects or effects of disturbances. Fundamental findings (neuron stimulability, extracellular signal characteristics) are also relevant to in-vivo or biomedical applications for improving recording quality, stimulation efficiency and safety. The methodology bears the potential to serve a broad range of applications, including biomedical applications, such as detailed investigations of the mechanisms of neurodegenerative diseases and applications in personalized medicine or pharmascreening of compounds with different types of electrogenic cells.

OBJECTIVES
The concise project objectives include:
• Study, at the same time, details of selected neurons or subcellular components (somas, axons, synapses, dendrites in a network context and the corresponding networks.
• Investigate, in detail, on the neuron/subcellular component side: (i) biophysical information that can be extracted from extra- and intracellular recordings, (ii) characteristics and roles of axons and axon initial segments (AIS), (iii) correlations between modulations in axonal signals and postsynaptic signals, and (iv) input-output functions of neurons and their input from presynaptic neurons.
• Investigate on the network side: (i) network architectures/organization (hub neurons, “rich clubs”) and functional connectivity, (ii) network development (short and long-term) and emergent phenomena, and (iii) reactions upon perturbations: network plasticity and homeostatic effects.
• Explore correlations between network size and function.
• Study concurrently interdependent component and network information processing characteristics.
• Study how component changes - over time or upon defined network perturbations - affect network behavior and how, in turn, network behavior or alterations in network behavior affect component characteristics.
• Use the obtained measurement results and data sets for modeling on different scales.
"WORK AND MAIN RESULTS
We started the activities in both subprojects 1 & 2 (components and networks). We used HD-MEA-technology for deciphering details of neuronal signaling at subcellular-compartment, individual-neuron and neuronal-network levels. Moreover, we combined upright confocal microscopy and patch clamp with HD-MEA technology and developed methods to evaluate complex multidimensional neuronal-activity data sets. Due to more recent scientific developments, we have also included brain organoids and human induced pluripotent-stem-cell-derived (hiPSC)-derived neurons as additional model systems into our ERC-AdG efforts, as the advent of those two neuronal preparations provides unique opportunities and may help to better recapitulate the behavior of neurons in mammalian or even human brains in healthy and diseased states. A number of scientific results have been published, as as can be seen in the Publication List.

We demonstrated that noise and signal attenuation depend on impedance rather than size of electrodes <10 μm diameter (Viswam Frontiers 2019). Further, we demonstrated the use of impedance spectroscopy (EIS) along with electrophysiology recordings on 59,760 electrodes at 13.5 μm spatial resolution (Viswam IEEE TBioCaS 2018). We were able to monitor position, adhesion, size, and electrical activity of brain slices.
We combined HD-MEAs with intracellular patch-clamp in order to realize a novel approach to precisely control the activity of presynaptic neurons (Jäckel Sci. Reps. 2017, Obien Sci. Reps. 2019). We evoked short- and long-term synaptic plasticity through manipulation of multiple synaptic inputs to a specific neuron. We used the combination patch clamp/HD-MEA technology for slice preparations and developed models for extracellular signal detection in slices. Moreover, we were able to study electrical-activity patterns of single neurons and neuronal networks in organotypic hippocampal slices cultures over several weeks.
We demonstrated precise microstimulation of single neurons by combining HD-MEA recording with immunostaining and confocal microscopy (Radivojevic Sci. Reps. 2016, Ronchi Frontiers 2019). We correlated morphological and electrical features of neuronal compartments with their responsiveness to extracellular stimulation. We developed strategies to assign electrical signals to axon initial segments (AIS), axonal arbors and proximal somatodendritic compartments. Stimulation at the AIS provided immediate, selective and reliable neuronal activation. Current stimulation proved to be more efficient (less than 2 pC) and caused less artifacts. AS HD-MEA stimulation can cause large electrical artifacts on neighboring electrodes, we developed a ""soft-reset"" technique, which shortened the artifact on electrodes neighboring the stimulation site to < 150μs (shorter than an AP) (Shadmani IEEE TBME 2018).
In rat cortical cultures and acute mouse slices, we found that the AIS dominates the extracellular AP landscape and the soma only contributes to a small extent (Bakkum Adv. Biosys. 2019). We showed that electrical activity initiates at distal end of AIS and spreads into axon proper and backward through the proximal AIS towards the soma as was evidenced through characteristic extracellular waveforms across different neuronal compartments.
We demonstrated noninvasive and direct recording of individual APs along millimeter-length axonal arbors in cortical cultures at microsecond temporal resolution (Lewandowska Frontiers 2016, Radivojevic eLife 2017, Emmenegger Frontiers 2019). We found that cortical axons conduct single APs with high temporal precision (~100 µs arrival time jitter per mm length) and reliability: in more than 8,000,000 recorded APs, we did not observe conduction- or branch-point failures.
We developed an automatic spike sorting algorithm, which can be scaled for HD-MEAs (Diggelmann J. Neurophys. 2018). The spike sorter is competitive with state-of-the-art spike sorters in terms of sensitivity and precision, while parameter adjustment or manual cluster curation is not required.
We developed a technique for accurate reconstruction of monosynaptic connections by combining HD-MEA and patch-clamp recordings. We used spike-triggered averaging of patch-clamp traces and strategies to deal with synchronous AP firing (Bartram Frontiers Conf. 2018). The mean postsynaptic potentials reliably identified connections, the amplitude of which can be used as measure of synaptic strength. We examined induction mechanisms of homeostatic synaptic plasticity in neuronal cultures. Results were presented at SfN 2018 with a journal article to follow.
We developed means to derive functional network connectivity from spike-train dynamics by probing short-latency interactions between pairs of neurons. We derived single-unit activity from HD-MEA network recording and modified the spike-sorting pipeline for primary rodent cultures to deal with bursting, which hampers reliable unit inference. The resulting spike-sorted data provided spike timing of several hundreds to thousands of units at sub-millisecond resolution. Combining waveform features and firing dynamics of spike-sorted units allowed us to classify neurons as excitatory or inhibitory and to infer cell-type specific connectivity. Putative excitatory monosynaptic connectivity inferred by pairwise cross-correlation analysis was sparse and comprised few high-degree hubs.


"
We progressed beyond the state of the art in several scientific and technological aspects as detailed below. Expected results are also briefly sketched.

Elucidation of the prominent role of the axonal initial segment (AIS)
By recording voltages of single neurons in dissociated rat cortical cultures and Purkinje cells in acute mouse cerebellar slices through hundreds of densely packed electrodes, we found (Bakkum Adv. Biosys. 2019) that the axon initial segment dominates the measured extracellular-action-potential landscape, and – surprisingly - the soma only contributes to a minor extent. The recorded dominant signal has negative polarity (charge entering the cell) and initiates at the distal end of the AIS. Interestingly, signals with positive polarity (charge exiting the cell) occur near some but not all dendritic branches and occur after a delay. Such basic knowledge about which neuronal compartments contribute to the extracellular voltage landscape is important for interpreting results from all electrical readout schemes.
Moreover, the AIS was found to be also the most stimulable compartment of a neuron (Radivojevic Sci. Reps. 2017). This knowledge is crucial for precise stimulation of individual targeted neurons to study single-neuron and neuronal-network characteristics (Ronchi Frontiers 2019). We could stimulate axon initial segments AISs with charges of less than 2 pC, which resulted in minimal artifact production and reliable readout of stimulation efficiency directly at the soma of the stimulated cell.

Tracking of individual single action potentials along axons
Axons are neuronal processes specialized for conduction of action potentials (APs). The timing and temporal precision of APs when they reach each of the synapses are fundamentally important for information processing in the brain. Due to small diameters of axons, direct recording of single AP transmission is challenging. Consequently, most knowledge about axonal conductance derives from modeling studies or indirect measurements. We demonstrated a method to noninvasively and directly record individual APs propagating along millimeter-length axonal arbors in cortical cultures with hundreds of microelectrodes at microsecond temporal resolution (Radivojevic eLife 2017). We found that cortical axons conduct single APs with high temporal precision (~100 µs arrival time jitter per mm length) and reliability: in more than 8,000,000 recorded APs, we did not observe any conduction or branch-point failures. Upon high-frequency stimulation at 100 Hz, successive APs became slower, and their arrival time precision decreased by 20% and 12% for the 100th AP, respectively.

Network Characterization
We developed means to derive functional network connectivity from spike-train dynamics by probing short-latency interactions between pairs of neurons. We derived single-unit activity from HD-MEA network recording and modified the spike-sorting pipeline for primary rodent cultures to deal with bursting, which hampers reliable unit inference. The resulting spike-sorted data provided spike timing of several hundreds to thousands of units at sub-millisecond resolution. Combining waveform features and firing dynamics of spike-sorted units allowed us to classify neurons as excitatory or inhibitory and to infer cell-type specific connectivity. Putative excitatory monosynaptic connectivity inferred by pairwise cross-correlation analysis was sparse and comprised few high-degree hubs. This line of research will yield more results in the 2nd half of the project.

Automatic spike sorting
We developed an automated approach to spike sorting, the assignment of electrical activity to units or neurons. Most existing spike sorting methods for single electrodes or small multielectrode arrays suffer from the “curse of dimensionality” and cannot be directly applied to recordings with hundreds of electrodes. This holds particularly true for the standard reference spike sorting algorithm, principal component analysis-based feature extraction, followed by k-means or expectation maximization clustering, against which most spike sorters are evaluated. We developed a spike sorting algorithm (Diggelmann J. Neurophys. 2018) that circumvents the dimensionality problem by sorting local groups of electrodes independently with classical spike sorting approaches. It is scalable to any number of recording electrodes and well suited for parallel computing. The combination of data prewhitening before the principal component analysis-based extraction and a parameter-free clustering algorithm obviated the need for parameter adjustments.

Combinations of techniques
We successfully combined whole-cell patch-clamp recordings with HD-MEAs. Simultaneous intra- and extra-cellular recordings allowed us to manipulate somatic membrane potentials, while monitoring the extracellular axonal action potentials throughout the axonal arbor. First experiments were conducted to induce pharmacological and sub-threshold modulation of membrane potentials. In addition, we built a combined HD-MEA - dynamic clamp setup, which facilitates the study of voltage-gated channels by simulating a particular conductance. More results are expected for the next project period.
Dendritic properties govern the interaction of synaptic inputs, which gives rise to a diverse repertoire of localized input processing options. To facilitate the investigation of such input interactions, we developed an approach that includes the combination of HD-MEAs with simultaneously applicable upright confocal microscopy, which provides, for localized synapses, precise activation sequences of the respective presynaptic cells, while it also allows for imaging the evoked dendritic signals. To this end, we performed simultaneous high-density microelectrode array (HD-MEA) recordings and Ca2+/voltage imaging in primary neuronal cultures. By correlating unit activity with Ca2+ signals in dendritic spines, we were able to associate synapses with the corresponding presynaptic cells. First results have been submitted to the 2019 SfN conference, more results are expected for the second half of the project.
Moreover, we combined HD-MEA recordings with single-cell RNA seq to study how changes in gene expression (of single cells) correlate with the observed changes in electrical activity of neurons and their connectivity after a perturbation experiment. We think that a combined analysis of neuronal activity, synaptic connectivity and gene expression patterns is a promising approach to study the mechanisms underlying neuronal homeostasis. First results have been submitted to the 2019 SfN conference and more results are expected in the next 2 years.

Organoids
We established a protocol to grow embryonic-stem-cell derived 3D-neuronal tissue on HD-MEAs. These so-called brain organoids were then used to track the development of neuronal networks and study their emerging electrical activity and connectivity. First results were presented at the 2018 MEA meeting (Schröter Frontiers Conf. 2018 and Girr SfN 2018), a manuscript is in preparation.
EM of neurons on chip, false colors
Confocal image of motorneurons on a chip